Relation between voxel-based specific regional analysis system for Alzheimer's disease (VSRAD) on 3-tesla MRI and...

Voxel-based specific regional analysis system for Alzheimer's disease (VSRAD) software using MRI scanner allows quantification of hippocampal and parahippocampal atrophy in the medial temporal structures by Z-score, and this score is widely used in clinical Alzheimer's disease (AD) diagnosis. However, it is unclear whether the Z-score is useful to discriminate normal aging from cognitive impairment (CI) or mild cognitive impairment (MCI). The present study examined the associations between VSRAD Z-score and cognitive performance quantified by Memory Performance Index (MPI) and determined a Z-score cut-off value.

Three-tesla brain MRI was conducted in 100 outpatients without dementia, and all MRI data were analyzed using VSRAD. The target region of interest (ROI) mainly consisted of the para hippocampal gyrus. The degree of atrophy in the ROI was obtained from the averaged positive Z-score of the ROI. Cognitive performance was evaluated with the Japanese version of the MCI screen (MCIS). Patients were classified into normal (NL) and below normal (BNL) cognitive groups by MPI. The relation between MPI and VSRAD Z-score were assessed with logistic regression analysis, and the cut-off value for Z-score was determined by receiver operating characteristic curve analysis.

Sixty-two percent (62%) were identified as the BNL group by MPI. Univariate analyses found that the BNL group had a significantly higher age, shorter years of education, and higher Z-score in VSRAD compared to the NL group, but no statistically significant difference was observed between genders. Bivariate correlation found that MPI, which is adjusted for age, gender, and years of education, was significantly correlated with Z-score assessed by VSRAD (Pearson's r = -0.52, p < .001). A subsequent logistic regression of VSRAD Z-score on BNL classification was used to generate a receiver operating characteristic curve (AUC = 0.75). The Youden index was applied to identify a cut-off value of VSRAD Z-score of 1.14 (sensitivity = 62.9%; specificity = 84.2%) to classify MPI < 50.2 (BNL) with overall accuracy of 73.5%.

VSRAD Z-score using VSRAD software was one independent factor significantly associated with cognitive performance measured by MPI. The determination of a cut-off value for Z-score (1.14) that can help discriminate normal patients from those with MCI.